liana.mt.estimate_metalinks

liana.mt.estimate_metalinks(adata, resource, pd_net, t_net=None, x_name='metabolite', y_name='receptor', use_raw=False, layer=None, verbose=False, **kwargs)

Estimate Metabolites from anndata object, and return a MuData object of metabolites and receptors.

Parameters:

adata AnnData

Annotated data matrix.

resource DataFrame

Resource to use for ligand-receptor inference.

pd_net DataFrame

Metabolic production-degradation network to use.

t_net DataFrame | None (default: None)

Transport set to use.

x_name str (default: 'metabolite')

Name of the metabolite modality.

y_name str (default: 'receptor')

Name of the receptor modality. Must be present as a column in the resource.

use_raw bool (default: False)

Whether to use the .raw attribute of adata. Defaults to False (uses .X).

layer str | None (default: None)

Layer in anndata.AnnData.layers to use. If None, use anndata.AnnData.X.

verbose bool (default: False)

Verbosity flag.

**kwargs object

Additional arguments to pass to the decoupler-py functions. Method-specific arguments are not supported.

Return type:

MuData

Returns:

A MuData object with metabolite & receptor assays.

Examples

pd_net is a metabolite-to-enzyme network whose weights say whether a gene produces (+) or degrades (-) a metabolite, and resource links metabolites to their receptors. Both normally come from MetalinksDB (see liana.rs.get_metalinks()); toy ones are built here so the example stays offline:

>>> import numpy as np
>>> import pandas as pd
>>> import liana as li
>>> adata = li.ds.generate_toy_adata()
>>> genes = adata.var_names[:16].tolist()
>>> pd_net = pd.DataFrame({"source": np.repeat(["HMDB0000122", "HMDB0000148"], 8), "target": genes, "weight": 1.0})
>>> resource = pd.DataFrame({"source": ["HMDB0000122", "HMDB0000148"], "receptor": ["CD4", "ITGB2"]})
>>> mdata = li.mt.estimate_metalinks(adata, resource=resource, pd_net=pd_net)

Metabolite abundances are estimated from the enzyme expression and returned in a 'metabolite' modality, next to the receptors in a 'receptor' one. Pass t_net to additionally require a transporter for metabolites that cannot cross the membrane on their own. The result is the input to liana.mt.bivariate or to any single-cell method, with x_mod='metabolite' and y_mod='receptor'.